Switch ifcmp+brif folding to be icmp+brz folding, which is what frontends actually produce.

cranelift/filetests/filetests/simple_preopt/branch.clif

@@ -1,31 +1,58 @@
 test simple_preopt
 target x86_64
 
-function %brif_to_brz_fold(i32) -> i32 {
+function %icmp_to_brz_fold(i32) -> i32 {
 ebb0(v0: i32):
-    v1 = ifcmp_imm v0, 0
-    brif eq v1, ebb1
+    v1 = icmp_imm eq v0, 0
+    brnz v1, ebb1
     jump ebb2
 ebb1:
-    v2 = iconst.i32 1
-    return v2
-ebb2:
-    v3 = iconst.i32 2
+    v3 = iconst.i32 1
     return v3
+ebb2:
+    v4 = iconst.i32 2
+    return v4
 }
-; sameln: function %brif_to_brz_fold
+; sameln: function %icmp_to_brz_fold
 ; nextln: ebb0(v0: i32):
-; nextln:     v1 = ifcmp_imm v0, 0
+; nextln:     v1 = icmp_imm eq v0, 0
 ; nextln:     brnz v0, ebb2
 ; nextln:     fallthrough ebb1
 ; nextln: 
 ; nextln: ebb1:
-; nextln:     v2 = iconst.i32 1
-; nextln:     return v2
+; nextln:     v3 = iconst.i32 1
+; nextln:     return v3
 ; nextln: 
 ; nextln: ebb2:
-; nextln:     v3 = iconst.i32 2
+; nextln:     v4 = iconst.i32 2
+; nextln:     return v4
+; nextln: }
+
+function %icmp_to_brz_inverted_fold(i32) -> i32 {
+ebb0(v0: i32):
+    v1 = icmp_imm ne v0, 0
+    brz v1, ebb1
+    jump ebb2
+ebb1:
+    v3 = iconst.i32 1
+    return v3
+ebb2:
+    v4 = iconst.i32 2
+    return v4
+}
+; sameln: function %icmp_to_brz_inve
+; nextln: ebb0(v0: i32):
+; nextln:     v1 = icmp_imm ne v0, 0
+; nextln:     brnz v0, ebb2
+; nextln:     fallthrough ebb1
+; nextln: 
+; nextln: ebb1:
+; nextln:     v3 = iconst.i32 1
 ; nextln:     return v3
+; nextln: 
+; nextln: ebb2:
+; nextln:     v4 = iconst.i32 2
+; nextln:     return v4
 ; nextln: }
 
 function %brif_inversion(i32) -> i32 {